Depth covariance computation

[JzHuai0108]

Dear author,

I am confused by the lines for depth covariance computation quoted below,

                F = torch.matmul(Q_ * E_sum.t(), L_inv) # K*HW x D*P
                F2 = torch.pow(F, 2)
                delta_cov = F2.sum(dim=-1) # K*HW

My expected form is like below according to your probabilistic fusion paper eq 5,

F = torch.matmul(Q_ * E_sum.t(), L_inv.t()) # K*HW x D*P

Note I think there is a transpose for L_inv. Can you please clarify this?